Prediction of fracture toughness parameters of epoxy-carbon fabric-CNT composites using Taguchi's and machine learning approach

ABSTRACT This study investigates the prediction and optimization of fracture toughness parameters of carbon fabric/epoxy composites enhanced with carbon nanotubes (CNTs). The composites were fabricated by hand layup method, and the effects of filler addition, crack length/specimen width ratio (a/W), and temperature were examined. Taguchi’s method was used for experimental optimization, while Random Forest was applied as the predictive modelling technique. Results revealed that fracture toughness improved with CNT addition, reaching optimum performance at 0.5 wt.% CNT, a/W ratio of 0.55, and room temperature (25 °C). The Taguchi analysis confirmed filler addition as the most influential factor, followed by temperature and a/W ratio. ANOVA indicated filler addition had the maximum influence, followed by temperature and a/W ratio. Comparative performance showed that Random Forest outperformed Taguchi’s method, with a lower mean squared error (MSE = 1.156 vs. 3.009), highlighting the superior predictive capability of machine learning. The results contribute to develop the composite materials with better fracture toughness for industrial and structural applications.